Wednesday, July 15, 2015

Trump and Hispanics

So an ABC News/Washington Post poll came out today that received wide attention. Here's the lede:
Positive views of Donald Trump have doubled since his controversial comments on immigration, although many more Americans still dislike rather than like him, now by a nearly 2-to-1 margin -- and his negatives have soared among Hispanics, a sought-after group in national politics.

Let's think about this for a moment. It's a poll of 1,011 adults with a margin of error of 3.5 percent. Lemme be clear -- that margin of error does not attach itself to Hispanics or any other subgroup. No, that's for the entire sample of adults. Not Republicans. Adults. You have to do the math again for a subgroup based on how many you surveyed.

So how many Hispanics are in the survey? Hell if I know. The report doesn't tell us. Let's assume 100 Hispanics, and that's a big assumption. The margin of error for that subset is 10 percent. In other words, Trump's 81 percent negative could be as low as 71 and as high as 91, but we don't really know. Subsamples are tricky, especially Hispanic subsamples. To be safe, I'd put the margin of error at more like 20 percent, which makes the sample all but useless for analysis purposes. Instead, you should oversample Hispanics (or any other group) to really look at the data.


Mark E. Johnson said...

Yeah, but that would require work and, you know, math ...

Hollander said...

Yeah, there was another poll out today with many of the same issues, but that horse is beaten enough.